Hugues Thomas
University of Toronto, Apple (United States), Apple (United Kingdom)
Papers
7
Total Citations
58
H-Index
5
About
Hugues Thomas is a robotics researcher whose work lies at the intersection of machine learning, dynamic scene understanding, and safe autonomous navigation. His primary research focuses on enabling mobile robots to operate intelligently in complex, human-filled environments by predicting and reacting to future motion. Thomas’s major contributions include pioneering the development of Spatiotemporal Occupancy Grid Maps (SOGMs), a novel representation that embeds future information about dynamic scenes into a single, learnable map. His work on self-supervised and unsupervised learning methods allows robots to generate these predictive maps from noisy, real-world lidar data, eliminating the need for costly manual annotation. His 2022 paper on SOGMs for lifelong navigation has garnered 18 citations, while his 2021 work on unsupervised lidar feature learning for probabilistic trajectory estimation has accumulated 15 citations. Notably, Thomas has also advanced practical navigation systems, including SAFER, a collision avoidance system that combines reinforcement learning with real-world deployment to enhance safety, and DR-MPC, a deep residual model predictive control approach for social navigation. His research is distinguished by its focus on bridging the gap between simulation-trained models and the unpredictable realities of human motion, making his work highly relevant for the next generation of socially-aware robots.
Research Focus
Key Achievements
Top Papers
- 1
- 2
- 3
- 4
- 5
- 6
- 7